Explainer
Can a PLC Servo Valve Stop a Hydraulic Ram Repeatably
Short answer
No — not on a PLC analog output alone. Oil compressibility, load and valve null drift move an open-loop stop, so the loop must close on a ram-mounted transducer, which a Delta Motion note says can position heavy loads to better than 0.0005 in. Decide by measuring today's stopping spread against the part tolerance: if the stop switch meets it, modernize the servo positioner and keep the manual fallback.
On this page
Not on a a PLC analog output alone. A servo valve commanded open-loop stops a 30-ton ram wherever oil compressibility, load and spool null drift leave it, and those change with every part and every degree of oil temperature. Repeatable stops come from closing the loop on a ram-mounted position transducer and trimming the valve command until measured position equals target; a Delta Motion technical note puts a well-designed closed-loop hydraulic axis at better than 0.0005 in on heavy loads. Before any hardware choice, measure the stopping spread the existing servo-driven stop switch already gives and compare it to the part tolerance. If it fits, modernize the positioner; if it does not, build the closed-loop axis.
Why an open-loop valve command misses the stop #
Three physical effects move the stop, and none of them appears in the commanded signal:
- The trapped oil acts as a spring in series with the ram mass, so the ram keeps moving while pressure builds. Deceleration takes distance, and that distance changes with load.
- Spool flow depends on the pressure drop across the valve. The same milliamp command gives a different ram velocity at full bend force than on an unloaded approach.
- Spool null shifts with oil temperature, so a zero command is not quite zero flow on a cold morning, and not the same near-zero at working temperature.
A scaled analog output sees none of this, which is why the loop, not the actuator type, delivers the accuracy. Production bending machines that use hydraulics pair the servo valve with a position feedback loop, exactly as motorized versions pair a servo motor with a ball screw.
The two architectures that do work #
| Servo ball screw plus stop switch (today's machine) | Servo valve with closed loop on the ram | |
|---|---|---|
| What sets the stop accuracy | Switch repeatability, screw error, ram overrun after the trip | Transducer resolution and linearity, valve resolution, loop gain |
| Feedback | Encoder or resolver on the screw motor | Absolute transducer mounted on the ram |
| If the positioner fails | Set the switch by hand and run parts | Axis fault stops the machine |
| Heat added to the oil | Low: the positioner carries no process force | Throttling loss, pressure drop times flow, every cycle |
| Sizing inputs | Set points and tolerance from the part drawing | Valve rated flow at its stated pressure drop; ram area from the nameplate |
The existing concept is not a jumble; it is one of the two sound architectures. Its virtue is that the positioner sits outside the force path: the servo parks a switch, the ram runs until it trips, and a dead drive still leaves an operator who can set the stop by hand. The closed-loop axis buys accuracy by moving positioning inside the ram motion and gives up that fallback. That trade is the real decision, and it turns on a measurement, not on preference.
The natural frequency sets the loop you can run #
The cylinder-and-load combination has a hydraulic natural frequency, and it caps how fast the position loop can act before the ram oscillates. For a symmetric double-acting cylinder near mid-stroke, the standard approximation is:
ωn = √(4·β·A² / (V·M))
with β the effective bulk modulus in lbf/in², A the ram area in in², V the total trapped oil volume in in³ (area times stroke, both sides), and M the moving mass in lb·s²/in — weight divided by 386 in/s².
Worked example, calculated for these inputs, not measured:
- 30 tons ≈ 60,000 lbf at 3,000 psi needs A = 60,000 / 3,000 = 20 in², so a 5 in bore — the same force-over-area arithmetic as sizing a hydraulic cylinder bore.
- A = π/4 × 5² = 19.6 in²
- A 20 in stroke traps V = 19.6 × 20 ≈ 393 in³
- Ram and tooling weighing 2,000 lb give M = 2,000 / 386 = 5.18 lb·s²/in
- β = 150,000 psi, a typical effective value for mineral oil; entrained air lowers it
ωn = √(4 × 150,000 × 19.6² / (393 × 5.18)) = √(230,500,000 / 2,035) ≈ 337 rad/s, or fn = 337 / (2π) ≈ 54 Hz.
Two rules of thumb, labelled as such: run the closed-loop bandwidth at no more than about a third of fn, here about 18 Hz; and sample at least ten times the loop bandwidth, here 180 Hz or faster — an update every 5.5 ms or better. A longer stroke or heavier ram lowers fn with the square root of 1/(V·M), and every number derived from it shrinks with it.
Where the loop lives #
The PLC can write set points, sequence the machine and monitor faults. The loop itself wants a deterministic 1 to 2 ms update against the roughly 5.5 ms ceiling from the example. A regular scan of 10 to 50 ms with cycle-to-cycle jitter under-samples an 18 Hz loop, and the variable delay reads as phase lag that shows up as oscillation at the stop point. The decision: close the position loop in a dedicated motion or axis module, or in the valve amplifier's internal controller, both reading the ram transducer directly, and let the PLC supervise. A servo valve driven by a PLC is real only in that split form — supervisor in the scan, loop in the module.
Decide by measuring the stopping spread you have now #
Put a dial indicator or linear scale on the ram or jaw, command five set points across the stroke, cycle each repeatedly, and record the stop positions at cold oil and at working temperature. The spread — worst stop minus best stop at one set point — compared with the part tolerance is the whole decision:
- A total spread of 0.016 in (±0.008 in about the mean) against a ±0.020 in tolerance, illustrative numbers: the mechanical arrangement already meets the print. The failing item is the 17-year-old DC servo drive and motor, and the fix is a modern servo motor, drive and encoder turning the same ball screw with the stop switch untouched — the drive swap walked through in selecting a replacement servo drive for resolver motors is that job.
- A total spread of 0.080 in against a ±0.010 in tolerance: switch repeatability plus ram overrun cannot close that gap. Build the closed-loop axis: an absolute transducer on the ram with resolution a fraction of the tolerance, a servo or proportional valve matched to ram area and required velocity, and the loop closed in a module or amplifier.
Both paths keep an independent mechanical limit so a feedback fault cannot overstroke the ram.
Commissioning and proving the stop #
- Set the spool null first: ram unloaded, zero command, no creep.
- Close the position loop at low proportional gain and raise it until the stop just overshoots, then back off. Tune loaded: a 30-ton bend changes the natural frequency the loop sees.
- Command a deceleration ramp into the target instead of a step stop.
- Test both ends of the product mix: bend force reverses on the ram and changes with material and angle, so run the heaviest and lightest parts.
- Prove repeatability with the same five set points, recording the transducer and an independent dial indicator, cold and hot. The spread you measure is the accuracy; nothing on a datasheet substitutes for it.
- Log tank temperature across the runs. A throttling valve turns pressure drop times flow into heat, and a temperature that climbs through repeated cycles says the valve or the pressure setting is oversized for the duty.
- Hold the oil to the cleanliness class the valve datasheet specifies. A drifting spool from contaminated oil appears as position error that grows over weeks.
One safety boundary #
ISO 4413 sets the general safety requirements for hydraulic systems: stop commands take priority over motion commands, and any function relied on to prevent hazardous motion must be engineered and validated to the reliability the machine's risk assessment assigns it. The repeatable stop built here is a positioning function, not a safety function — guarding and emergency stop must bring the ram to rest whether or not the loop is healthy, so choose a valve whose de-energized spool state halts the ram, the same fail-safe logic that keeps a safety shutoff valve closed when its coil loses power, and keep the mechanical limit in place.
Questions asked
How fine should the ram position transducer be?
Use a rule of thumb of ten times finer than the stop tolerance, so quantization is noise rather than error: for a ±0.010 in tolerance, 0.002 in resolution or better. Mount it on the ram, not the screw, because jaw and linkage compliance never show up in a screw-mounted reading. Take the resolution and update rate from the transducer datasheet.
Can the PLC close the position loop in its regular scan?
Only if the scan is short and deterministic. The worked example allows about 18 Hz of loop bandwidth from a 54 Hz natural frequency, which wants samples at 180 Hz or faster — every 5.5 ms at worst, with the 1 to 2 ms range preferred. Most scans run 10 to 50 ms and vary cycle to cycle, so put the loop in a motion module or the valve amplifier and leave sequencing to the PLC.
What happens if the ram transducer fails mid-cycle?
The controller loses the position it regulates, and the ram can drive to a stroke end or coast uncontrolled. Monitor transducer validity continuously, drive the valve to its de-energized stop state on fault, and rely on the independent mechanical limit to prevent overstroke. Prove this failure mode at reduced pressure before production parts run.
Should the mechanical stop switch stay after a servo valve retrofit?
Yes. It remains the hard limit and the manual fallback: because the positioner sits outside the force path, an operator can set the stop by hand and run parts when the drive or axis is down. Folding positioning into the ram motion removes that fallback, so give it up only for accuracy the measured spread proves you need.
How repeatable can a closed-loop hydraulic stop actually be?
Better than 0.0005 in on a well-designed system, per a Delta Motion technical note — the ceiling of the architecture with a stiff valve, fast loop and rigid mounting. A retrofit earns only what its measured spread over repeated cycles shows, verified cold and hot. Treat that measured spread as the machine's accuracy number.